The execution time of an operator can not be changed by simply changing the start_date

[WesleyBatista]

Dear Airflow Maintainers,

Before I tell you about my issue, let me describe my environment:

Environment

• Version of Airflow (e.g. a release version, running your own fork, running off master -- provide a git log snippet) : v1.5.1
• Operating System: Linux airflow 3.16.0-4-amd64 Improving the search functionality in the graph view #1 SMP Debian 3.16.7-ckt11-1+deb8u5 (2015-10-09) x86_64 GNU/Linux
• Python Version: Python 2.7.9

Now that you know a little about me, let me tell you about the issue I am having:

Description of Issue

The scheduler do not execute the task considering the start_date param on the operator level (in this case a PythonOperator).

• What did you expect to happen?
  The task being scheduled on a different time.
• What happened instead?
  on the task details I can see that the start_date changed on the PythonOperator, however the executions prints on the log the time set on default_args
• Here is how you can reproduce this issue on your machine:

from airflow import DAG
from airflow.operators import PythonOperator
from datetime import datetime, timedelta

default_args = {
    'owner': 'airflow',
    'start_date': datetime(2016,03,21, 9, 0, 0),
    'depends_on_past': False,
    'wait_for_downstream': True,
    'sla': timedelta(minutes=30)
}
dag = DAG(
    'sample_dag',
    default_args=default_args,
    schedule_interval=timedelta(hours=24)
)

def python_callable(**context):
    return True

taskParameters = {
    "dag": dag,
    "task_id": "send_email",
    "python_callable": python_callable,
    "provide_context": True,
    # 'start_date': datetime(2016,03,22, 12,0,0),
}

task = PythonOperator(**taskParameters)

Reproduction Steps

1. I have start_date set on default_args on the DAG definition.
2. Leave running to get some successful tasks
3. I set a different start_date to the PythonOperator.
4. Clear some of the last successful executions to force the scheduler rerun the task
5. On the task log I get the following: [2016-03-28 09:00:50,997] {models.py:974} INFO - Executing <Task(PythonOperator): send_email> on 2016-03-26 09:00:00

I guess this happens due to the last execution date + schedule interval rule.

I think that the common expectation here is: after the change on the start_date, the new schedules to be executed are considering the new date.
But I do not know too if it goes against what the project proposes.

What do you guys think about it?